As telecom operators face rising complexity, mounting cost pressures, and relentless service demand, the industry is moving decisively beyond traditional automation toward something more ambitious: fully autonomous, self-optimizing networks. A recent session moderated by Abe Nejad of The Network Media Group (NMG) explored how operators and ecosystem partners are progressing from automation to autonomy and the role AI-driven intelligence across the RAN and core must play to make closed-loop operations a reality at scale.
Speakers:
Watch the full interview.
The panel opened with a grounding question: what does Level 4 autonomy actually mean in practice? The speakers put it plainly – it is the shift from human-assisted operations to agent- and system-assisted operations. For companies like Telenor, that definition has a very specific commercial dimension. Level 4 and above is about making the network more programmable so that it is easier to manage and, more importantly, more monetizable. The goal is not autonomy for its own sake. It is programmability in service of growth.
To demonstrate the scale of that ambition, Telenor has geographically ring-fenced a flagship initiative: a full Level 4 autonomous city in the Nordics. Rather than achieving autonomy in a single domain or a handful of nodes, the target is Level 4 everywhere within that boundary – a proof point designed to scale, not just to impress.
Both speakers were unequivocal on one point: autonomous networks are impossible without the openness that cloud-native architectures and Open RAN enable. Without true virtualization of the stack combined with the openness of Open RAN and RIC platforms, an autonomous network was never going to be possible. Staying locked into legacy, proprietary silos – where every feature flows from a single vendor – forecloses the very programmability that autonomy requires.
When you disintegrate the traditional vendor silos, the RAN becomes programmable. rApps and xApps can come from the operator, from third parties, from the ecosystem – not just from one incumbent. The leaders emphasized that the network needs to become more programmable, and that programmability is what ultimately powers monetization.
Both speakers were candid about the industry's current shortcomings. Telcos sit on vast stores of network data – from radio and transport layers to CRM and fault management systems – but correlating and activating that data in a sovereign, secure way remains an unsolved problem. After all, there is no AI without data.
One of the more refreshing moments in the session was the frank acknowledgment that autonomous networks are not purely a technology problem. They are an organizational and mindset challenge that the industry has consistently underestimated. Telcos remain slow-moving and siloed – a structural reality that Open RAN architectures are now, for the first time, giving operators a genuine opportunity to disrupt. Achieving Level 4 at scale demands flatter hierarchies, cross-domain champions, and intent-driven operating models, not just upgraded infrastructure.
The benchmark Rakuten Mobile has set makes this concrete. Running a nationwide network in Japan with approximately 200 people, built greenfield on a fully virtualized, open stack, represents a reference point that legacy operators look at with a mixture of admiration and urgency. If a NOC that once ran on 120 people could be run with 10 or 15, that is proof that the needle has actually moved. The aspiration is not marginal OpEx improvement. It is about 50 to 80 percent efficiency gains that redeploy human talent from alarm-watching to revenue-generating work.
The industry needs to stop counting Level 4 certifications and start showing what Level 4 actually delivers – not minor, incremental improvements, but demonstrated, scalable needle-movement. If it cannot scale, it does not count.
"I'm almost envious when you say you're running the network with 200 people... if you can say my NOC was 120 people, now I'm running it with 10 or 15, you've actually been able to move the needle."
"We operate a massive network pan Japan with around 200 people, which is commendable, with a simple set of solutions or one stack in a complete, virtualized manner."
